MCP SuperAssistant Chrome Extension is a browser-based bridge that brings Model Context Protocol capabilities directly into web AI interfaces like ChatGPT, Google Gemini, Perplexity, and Grok. Designed for developers, researchers, and AI power users, the extension detects tool invocations directly in conversation streams and executes them via a local or remote proxy server. It connects web-based AI chat interfaces to external systems, file systems, and API integrations defined in standard MCP configurations. Users can review tool calls manually or automate execution and response insertion back into the chat window. By handling communication between the web browser and local tool environments via Server-Sent Events, WebSockets, or Streamable HTTP, the extension enables web chat interfaces to interact with local development environments, databases, and custom scripts without switching applications.
Category: AI & LLM Tooling
Tags: browser, chrome, extension, integration
Visit MCP SuperAssistant Chrome Extension
config.json file defining your target MCP servers, or use an existing config such as Claude Desktop's configuration file. 3. Start the proxy server using: npx -y @srbhptl39/mcp-superassistant-proxy@latest --config ./config.json --outputTransport sse 4. Open a supported AI interface like ChatGPT or Gemini in your browser to display the sidebar. 5. Click the connection status indicator, enter the proxy URL (default: http://localhost:3006/sse), and select SSE transport. 6. Click Connect to link the browser extension to your local MCP tools.Part of MCP Servers
The extension supports numerous web interfaces, including ChatGPT, Google Gemini, Perplexity, Grok, Google AI Studio, OpenRouter Chat, DeepSeek, T3 Chat, GitHub Copilot, Mistral AI, Kimi, Qwen Chat, and Z Chat.
The extension uses a companion local proxy server run via npx. The proxy reads standard MCP configuration files and communicates with the browser extension using Server-Sent Events, WebSockets, or Streamable HTTP transports.
Ensure the extension is active, insert the required MCP instruction prompt into the chat before messaging, verify the proxy server is running, and try turning off web search while enabling reasoning mode on supported models.
Yes, the project is open source and released under the MIT license on GitHub, allowing users to inspect the code, contribute, and report issues.